Drupal websites that dream of content improvement
In the Bringing Drupal into your DNA webinar, Dries Buytaert described a compelling future: “Overnight these agents can make changes for review in the morning.” This talk explores what it takes to make that vision a reality when AI is used at scale within Drupal.
Inspired by an idea to use Drupal AI to audit all 400 local authority council websites in the UK against popular user journeys, a research project has been running over the last nine months to achieve just that.
There are many tools in the Drupal AI toolbox that focus on empowering the editor with tools they can click, orchestrate and execute. This project headed under the hood at how to orchestrate over 50,000 AI chat requests across thousands of web pages from lots of different websites.
This session is a practical case study in what breaks, what scales, and what Drupal AI needs next if it is going to build the alluring idea that websites could identify across a large editorial estate the issues that need fixing, ready for human review in the morning.
I will walk through the technical evolution of the system: starting with AI Automators and batch jobs, moving to cron-based execution, and eventually redesigning the architecture around concurrent workers using RabbitMQ & MinIO to achieve enterprise scale. Along the way, I encountered real-world challenges, including AI provider rate limits, fault tolerance, auditability, and working with different AI models. To address this, I implemented worker pools and a highly scalable library data storage system that enables concurrent execution across multiple locally hosted AI servers.
The talk will highlight why large-scale AI usage in Drupal requires new core capabilities: robust job orchestration, improved audit trails, concurrent worker execution, and better failure handling. These needs are not hypothetical—an editor managing thousands of pages will face similar demands when running overnight AI-driven analysis and recommendations.
This session is aimed at Drupal architects, contributors, and enterprise builders who want to understand what “AI at scale” really looks like. It also proposes how real, live case studies can help position Drupal AI as a credible enterprise platform—powered by Drupal, and ready for what’s next.
As an exciting climax the talk will end with an opportunity for everyone in the audience to perform a real time on-demand audit of their own website and key user journey.